Merge 8477d65 into 21da0c0

autokeras/preprocessor.py

@@ -263,13 +263,13 @@ def _transform(self, compose_list, data, targets):
 class DataTransformerMlp(DataTransformer):
     def __init__(self, data):
         super().__init__()
-        self.max_val = data.max()
-        data = data / self.max_val
-        self.mean = np.mean(data, axis=0, keepdims=True).flatten()
-        self.std = np.std(data, axis=0, keepdims=True).flatten()
+        self.mean = np.mean(data, axis=0)
+        self.std = np.std(data, axis=0)
 
     def transform_train(self, data, targets=None, batch_size=None):
-        dataset = self._transform([Normalize(torch.Tensor(self.mean), torch.Tensor(self.std))], data, targets)
+        data = (data - self.mean) / self.std
+        data = np.nan_to_num(data)
+        dataset = self._transform([], data, targets)
 
         if batch_size is None:
             batch_size = Constant.MAX_BATCH_SIZE
@@ -281,7 +281,6 @@ def transform_test(self, data, target=None, batch_size=None):
         return self.transform_train(data, targets=target, batch_size=batch_size)
 
     def _transform(self, compose_list, data, targets):
-        data = data / self.max_val
         args = [0, len(data.shape) - 1] + list(range(1, len(data.shape) - 1))
         data = torch.Tensor(data.transpose(*args))
         data_transforms = Compose(compose_list)